1 paper · 1 filter
Yin Lin, Elaine Ang, Erkang Zhu +2
LLM agents increasingly take on long-running tasks whose history grows far beyond a single model context window. Existing approaches compress earlier interactions or extract select…